The American tax code is a complicated system that confounds even the most veteran policymakers. The tax structure has long been based on income, where top-tier earners pay more than those at the bottom. But some of the professions that create the greatest public good, such as teaching or medical research, offer little compensation relative to the social benefits they create. In a recent paper, Wharton business economics and public policy professor Benjamin Lockwood examines whether it is possible to use the tax structure to encourage people to take on careers that promote social good. It’s an intriguing question right now because the public debate around taxation is pulling policymakers in different directions. Lockwood joined Knowledge at Wharton to talk about his research and how the findings add to the debate. The paper, “Taxation and the Allocation of Talent,” was co-authored with Charles Nathanson, a professor of finance at Northwestern University, and E. Glen Weyl, a principal researcher at Microsoft Research New England and a visiting senior research scholar at Yale University.
An edited transcript of the conversation follows.
Knowledge at Wharton: What is the premise behind this paper?
Benjamin Lockwood: The idea behind the paper was motivated by looking at the debates that we see in policy and in the popular press about what the tax system should look like. Conventionally, when economists think about the optimal tax system, we think about trading off the benefits of targeting resources to lower-income people — the benefits of redistribution — against the efficiency costs. When you have higher taxes, those higher marginal tax rates reduce the benefits of economic growth or of being really innovative or of working really hard, so it’s thought that lower incentive makes people less inclined to go out and create economic growth and be innovators.
In that model, the key driver of optimal tax rates, especially at the top of the income distribution, has to do with the efficiency costs. The question is, if you increase tax rates a little bit, how much less do people work? How much less do high earners work? One of the things that we were struck by is when you look around at popular debates, that’s often not what you hear people talking about when they argue for higher or lower taxes on high earners. Instead, you hear a lot of discussion about what the rich are actually doing, whether they’re the job creators and innovators who are hiring lots of people, or whether they’re rent-seekers who are sitting back and leeching off the resources of the rest of society.
Our goal with this paper was to try to map those popular debates into a more formal, optimal tax model. We thought those debates are better captured by a model in which people can go into different industries, select different professions, and some of those professions have positive spillovers onto the rest of society. They create benefits that are higher than the private compensation that people receive, whereas other professions might generate benefits that are lower than the private compensation they receive. If you believe that, then you might want to use public policy to encourage people to go into those socially useful careers where they are generating lots of social benefits.
Knowledge at Wharton: Some professions obviously create a lot of social good, such as teaching or research. But there are a lot of gray areas, too. How did you figure out which careers create social good?
Lockwood: There’s a ton of uncertainty here. One of the things that makes it a little more nuanced is that it’s not just which careers are generating lots of benefits, but which careers are generating benefits that are relatively larger or smaller than the private compensation that people are receiving. Our paper does not attempt to estimate those spillovers or those externalities directly. Instead, we present a theoretical framework for how we can think about using taxes or other policies to encourage people to go into professions that do have such spillovers. Then we look to a broad swath of existing economic literature that does try to estimate those sorts of things. There are individual papers here and there that have attempted to estimate those sorts of spillovers in particular industries.
For example, there is a nice paper by [Harvard professor] Raj Chetty and co-authors that looks at the benefits created by public school teachers. They find that teachers who are very good at adding value to students’ test scores over the course of a year, those students often go on to have higher earnings, lower rates of teen pregnancy and other things that are beneficial in society. By using tax records and the like, they’re able to put some numbers on that. They find that if you were to take a teacher from around the bottom of the distribution of teacher quality and replace him with a teacher that’s in the middle of the distribution, that would raise the net present value of their students’ earnings. And if you add up those increases in earnings across a whole classroom for a year, you get an increase on the order of $250,000.
By replacing a sort of inferior teacher with a median one, that’s an increase in student salaries that’s way bigger than the salaries that the teachers themselves receive. That’s the sort of calculation that you can do to figure out the ratio between the social benefits that are being created by a given profession and the private compensation they’re receiving.
“Conventionally, when economists think about the optimal tax system, we think about trading off the benefits of targeting resources to lower-income people against the efficiency costs.”
Another example along these lines is a nice paper by Kevin Murphy and Robert Topel that looks at the benefits from medical research. There’s a thought that one of the benefits of medical research is to increase the length of people’s lives. You can do some calculations to figure out how much people are willing to pay for increased longevity through what’s known as “the value of a statistical life.” You do things like look at how much people were willing to pay for airbags when those first came out and how much extra safety airbags provided. By using that kind of revealed preference strategy to place a monetary value on the increased longevity from medical research, they find that the kind of uncaptured gains over the last generation or two are on the order of 25% of GDP growth. So, a really large amount of overall growth has come in this uncaptured dimension from medical research, which is much larger than the private compensation that medical researchers are paid. When we review the literature, those examples of primary school teachers and medical researchers are by far the largest spillovers that we find across any of the professions. But those are the kinds of studies that we draw on to calibrate these spillovers for a bunch of different professions. We can build those into our model to ask, “What would this imply about the optimal tax rate structure or other kinds of policies?”
Knowledge at Wharton: Are there indications that the current tax system encourages people toward jobs that don’t create as much social good?
Lockwood: Maybe. I should say a little bit more about what we think the optimal tax structure would look like, and then we can kind of compare that to what we see in practice.
If you were motivated by this idea that some professions or industries have these positive or negative spillovers, a kind of first pass that you might make at optimal policy would be to directly subsidize or tax different industries based on whether those benefits are positive or negative. We do a couple of calculations along those lines in the paper.
I think there are a few reasons you might be a little skeptical about that or wonder whether that sort of policy would work. If education was thought to have big, positive spillovers and finance was thought to have negative spillovers, then I can easily foresee lots of financial consultants reclassifying themselves as “educating their clients” about how finance works so they can qualify for that sort of subsidy.
Put differently, although it’s easy for us right now in the data to determine who works in what sectors or professions, that’s in part because there’s nothing at stake in honestly reporting what sector or profession you work in. But if we suddenly started subsidizing or taxing different professions based on those categorizations, there would be a much greater incentive to misreport or make things cloudy. Once those kinds of tax instruments are on the table for adjustment, you can imagine large lobbying groups springing up to try to push them one way or another, possibly away from whatever the economic research says those spillovers are.
Knowledge at Wharton: If we’re not going to tax based on profession, what’s another way of doing it?
Lockwood: Another possibility would be to note that in the existing economy, the distribution of income differs across different professions. People in finance have a pretty high distribution. People working as kindergarten teachers have a much lower distribution. That means tax rates that fall on high earners tend to fall more heavily on people working in finance, whereas the marginal tax rate affects people toward the lower end of the distribution. Put differently, if you increase marginal tax rates in the areas of the distribution between what kindergarten teachers tend to earn and what financiers tend to earn, you will lower take-home incomes for financiers and increase tax revenues while not lowering the take-home pay of kindergarten teachers. You could, in fact, use the increased tax revenues to subsidize people toward the bottom of the distribution.
All of this is to say that because there are different income distributions in different professions, we could potentially use the nonlinear tax code that we already have to relatively raise or lower the benefits to earning in different parts of the income distribution overall, and that would implicitly change the returns to going into one profession versus another. We also look at that as a possible instrument, where we say, “Suppose the government can’t directly target various professions but instead has the same kind of income tax that we currently have. What would that optimal nonlinear income tax look like if you were designing it to improve the allocation of talent across individuals?”
This was the backstory that I wanted to tell, the foundation I wanted to lay, to make sense of comparing what we find to be the optimal-looking tax relative to what we actually see in society. I think there are a couple of interesting things here. One, if you take these estimated spillovers from the literature at face value and run a computer program to figure out what the optimal tax rates are, you get something that’s not so different from the kinds of tax rates that we currently see. We found that tax rates at the top of the distribution would be somewhere around 40%. That’s assuming there’s no benefit at all from redistribution. This is just a tax code whose sole goal is to raise efficiency by allocating talented people into various professions. If on top of that you wanted to do additional redistribution, then that would further make the tax code more progressive. But just looking at this efficiency-maximizing tax code, you’d get something that looks a bit like what we already have.
“If there’s a negative or unfortunate implication of our paper, it’s that just adjusting the nonlinear tax code does not achieve very large targeted welfare gains.”
The flip side is that there would also be some benefits from specifically targeting professions, and it does seem like our existing tax code does a bit of that, too. There are certain professions that are more heavily subsidized than others or that are more frequently part of the public sector. For example, the National Institutes of Health provides lots of grant subsidies that effectively promote and incentivize basic medical research. Most public education is funded directly from government spending, so the salaries in those professions are largely set by the public sector. Those are some things that could directly be set and adjusted.
Knowledge at Wharton: What would be the optimal way of setting up the tax structure to achieve this goal that you’re trying to get at?
Lockwood: There are some things in the existing tax code that resemble the things we find to be optimal, where you directly target professions that are thought to have these positive spillovers. But a big difference is that the spillovers estimated in the literature are so large that it suggests that the optimal subsidies would actually be way bigger than the ones that we currently have.
I guess if there’s a negative or unfortunate implication of our paper, it’s that just adjusting the nonlinear tax code does not achieve very large targeted welfare gains. Although you are able to relatively increase the benefits of becoming a teacher versus a financier, there’s only so much you can do with the blunt instrument of the nonlinear income tax. Our estimates suggest that the much more targeted effects of specifically subsidizing some sectors would be very, very large. As I’ve already mentioned, there are some real downsides of trying to do that specific targeting. It opens the door to lobbying, and it might be quite difficult to have these general, profession-specific taxes. A more pragmatic policy that could get a lot of these gains while not opening the door to quite so much manipulation would be something more like the existing NIH subsidies or public-sector schooling, with the public sector determining teachers’ salaries but adjusting those targets to bring them more into line quantitatively with the large spillovers that are estimated in those industries. This would involve drastically increasing the amount of public funds that are spent on subsidizing university medical research, or drastically increasing the salaries that are spent on quality teachers to encourage really talented people to go into those professions.
Knowledge at Wharton: What is the takeaway for policymakers? It seems like this paper is proposing a different way of looking at the tax structure. We look at it on an income basis, and this suggests looking at it more by profession or based on social good.
Lockwood: Yes, that’s right. In some sense, this flips around a bit the way that economists often think about the existing tax structure. On the other hand, I noted at the beginning that one motivation of this paper was to try to bring our economic models of optimal taxation into alignment with the policy debates that we’re already seeing. I think that public debates have been out ahead of economists on this front a bit already. You often hear policymakers saying things like, “We want to encourage people who are thinking about going to Wall Street to instead go into STEM professions and create new inventions that will increase economic growth.” That’s an argument that’s in line with this sort of reasoning, that there are real benefits from having talented people go into these sectors that have beneficial spillovers for the rest of us.
I think the main lesson for policymakers would not be so much this qualitative insight that it’s helpful to direct talented people into professions with positive spillovers. Instead, it would be the more quantitative insight that when we do look to the economic studies on this, when we look at the economic literature, there are a few really big, standout spillover professions — primarily teaching and medical research. There are already some economic policy levers that we have at our disposal that could be used to try to encourage talented people to go into those things.
Knowledge at Wharton: Where will this research take you next?
Lockwood: I think the biggest challenge here is still that these estimates of the spillovers are pretty vague. We have taken what our reading suggests is the best estimates from the literature right now, and it looks like a few of those are quite large. But there are several sectors where there’s just really no research that successfully quantifies how big these spillovers are.
One of the most important ones right now is entrepreneurship and new business creation. That’s something I suspect has spillovers and would make sense to try to pin down in an economically rigorous way. But because we don’t have much information on that yet, we had to leave it out of the paper.
There are also other sectors that we think have more abstract spillover benefits that are hard to quantify. Art and music are things that have very large benefits, and thinking about how to incorporate those into this kind of framework is something that we punted on, but that I think would be interesting for future research.
Knowledge at Wharton: There has been a lot of discussion about trickle-down economics, which is the notion that people in the top tier will respond to tax breaks by using the extra money to create jobs for the bottom tier. But your research turns that on its ear a little bit by looking at how the spillover effect depends on the job rather than the income.
Lockwood: That’s exactly right. I don’t love the term trickle-down economics because I think it was developed as a pejorative way to talk about this one particular economic theory that having lower tax rates at the top would let people at the top of the distribution take home more income. Then the question is what they do with that income.
If you believe that what they do with that income is create new jobs and new businesses and employ lots of people, which in turn creates lots of economic growth and raises the well-being of people throughout the income distribution, then I think one way of interpreting that view is to say that the people at the top of the income distribution have these kinds of positive spillovers. When it’s reframed in that way, rather than this lightning rod term of trickle-down economics, it becomes much easier to think logically about and also to try to bring economic data to bear. We can ask, what is it? What are the jobs that are actually being done by people at the top of the income distribution? Is there evidence that those jobs create those kinds of spillovers for the rest of us?
Once we ask that of the top of the income distribution, there’s really no reason to stop there. We might as well ask for all sorts of points across the income distribution, what are the spillovers from the work that people are doing at that point in the income distribution? What are the spillovers to the rest of society?